No, and I would be wary of anyone who says yes. The models have different strengths, they change every few months, and the useful skill is knowing what to use and when.
On my own systems the division is explicit and written down. One frontier model researches, plans and drafts, because that is where its judgement is best. A different one runs the day: QA sweeps, audits, the morning briefing, because it is the more careful reader and it is useful to have a second opinion that was not trained by the same company. A third builds and maintains the machinery the agents run, in sessions with me. A local model on my own hardware does the volume, reading whole books and batches of text, with no authority to approve anything anywhere. And the paintings on this site come from an older image model I keep reaching for because nothing newer touches it for abstraction.
The same principle runs the scoping loop I use on my own product: scope with one model, hand the scope to another to find the holes, bring the findings back to the first to close them. Two models disagreeing is worth more than one agreeing with me.
What I would say is a favourite is the shape around the models: who may do what, how you know a job happened, and what refuses to run when the answer is wrong. That part does not change when the leaderboard does.
My own answers, from my own work. They change when the work does.